I also noticed that you wrote something about a Fairlight IIx and its voice
architecture (needing to load a sound into each "voice" separately); this
was only true for the Fairlight I; the Fairlight II (and upward) could load
voices into all 8 cards simultaneously (they still were monophonic, though).

The Fairlight III goes back to the one voice per card bit, however.  That is
what an owner has told me, correct?

No, that's not correct. Not really.

So it would be only the II and IIx that had true dynamic assignment, right?

Even the II and IIx did not have dynamic assignment. Let me explain:

The older Fairlights had sound memory on each voice card and no common
sound memory. This means that each voicecard was playing back the sound
that was loaded into it. With the Series I this meant that you had to load
a single sound into each voicecard separately if you wanted to play it
8-voice-polyphonic (the earlier Fairlights also had a single polyphonic
output in addition to the mono outputs on the voicecards). The II and IIx
could load a single sound into as many voicecards as you desired in one
step.

The Series III has a common memory scheme similar to most samplers. Still,
the single outputs of the voices are still monophonic and don't have
dynamic assignment. Enter 1989 the digital router: This new card gave the
Fairlight 24 polyphonic outputs with dynamic assignment (the machine was
still 16-voice, though).

--
Malte Rogacki
----------------------------------------------------------------------------
"Don't forget to TURN ON THE SYNTHESIZER. Often this is the reason why you
 get no sound out of it." (ARP 2600 Owner's Manual)
----------------------------------------------------------------------------